agentleFS
UnfoldCRO-OrganizationUnfoldCRO-Organization/cro-agent-skills/skills/build-review-loop/SKILL.md

build-review-loop

Skill0 starsChanged 22 days agoMITView on GitHub

Default build-and-review loop for any real coding request typed in the composer — writing, editing, refactoring or fixing application code, components, endpoints, scripts, migrations or config with logic. Runs plan → build → parallel review panel (Claude + Codex) → merged issue table → fix all P1s → re-review → summary. Auto-invoke on any non-trivial code change. SKIP for trivial one-liners, comments, docs, config typos, or when the user says "skip review".

  • Reads credentials

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run this loop for any request that writes or changes code with logic: features, edits, refactors, bug fixes, endpoints, components, …

    line 15
  2. Distinctive:

    Write the code. Follow project conventions, the active project's CLAUDE.md, and the locked rules (PR-only, no AI attribution, keep …

    line 29
  3. Very specific:

    Verify a provider is reachable: bash ~/.claude/review-panel/panel-review.sh <provider> --check (expect http=200).

    line 59

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Default Build & Review Loop

Task: $ARGUMENTS

Encodes the standing loop from the global and project CLAUDE.md. Run it automatically for any real coding request — no slash command needed. Stay in caveman mode throughout so output stays cheap (caveman already relaxes for the plan/summary and safety-critical text).

When to run

Run this loop for any request that writes or changes code with logic: features, edits, refactors, bug fixes, endpoints, components, scripts, migrations, non-trivial config.

Skip the loop (do the edit directly) for:

  • Trivial one-liners
  • Comments or docs
  • Config typos
  • The user explicitly says "skip review"

The loop

1. Plan

Reply with a 4–5 line summary/plan of the approach, then proceed. No essay — just the shape of the change and the files touched.

2. Build

Write the code. Follow project conventions, the active project's CLAUDE.md, and the locked rules (PR-only, no AI attribution, keep sensitive data local, respect deploy windows).

3. Review panel (parallel)

Get independent reviews of the diff. Active reviewers: Claude + Codex.

  • Claude — run /review (or a focused self-review of the diff).
  • Codex — run /codex review.
  • Kimi / DeepSeek — ⏸️ ON HOLD (no account balance). Do NOT call them until re-enabled in ~/.claude/CLAUDE.md.
  • Any active reviewer that exits non-zero / is unauthed → mark skipped and continue. Never block on one reviewer.
  • Log each active reviewer's result as a JSON line: gstack-review-log '{"skill":"build-review-loop","status":"...","findings":...}' (reuses the existing <branch>-reviews.jsonl.)

4. Issue table

Merge all findings into one table, dedup overlaps, note where reviewers agree:

#SeveritySource(s)File:LineIssueSuggested fix

5. Fix

Rewrite the code to resolve every P1.

6. Re-review

Repeat steps 3–5 until no P1 remains. Max 3 rounds. If a P1 survives after 3 rounds, report it honestly instead of looping forever.

7. Summary

Close with a 4–5 line wrap-up: what changed, what each model flagged, what's left.

Reviewer keys / verification

  • Kimi → ~/.config/moonshot/key (or MOONSHOT_API_KEY)
  • DeepSeek → ~/.config/deepseek/key (or DEEPSEEK_API_KEY)
  • Verify a provider is reachable: bash ~/.claude/review-panel/panel-review.sh <provider> --check (expect http=200).
  • To re-enable Kimi/DeepSeek: recharge the account, confirm the --check returns http=200, then add them back as active reviewers in ~/.claude/CLAUDE.md and here in step 3.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.